Trends in Hit-to-Lead Optimization Following DNA-Encoded Library Screens

Abstract

DNA-encoded library (DEL) screens have emerged as a powerful hit-finding tool for a number of biological targets. In this Innovations article, we review published hit-to-lead optimization studies following DEL screens. Trends in molecular property changes from hit to lead are identified, and specific optimization tactics are exemplified in case studies. Across the studies, physicochemical property and structural changes post-DEL screening are similar to those which occur during hit-to-lead optimization following high throughputscreens (HTS). However, unique aspects of DEL—the combinatorial synthetic methods which enable DEL synthesis and the linker effects at the DNA attachment point—impact the strategies and outcomes of hit-to-lead optimizations.

The discovery of hits and their development into tractable chemical leads is critical to chemical biology and drug discovery.¹ DNA-encoded library (DEL) screens offer the ability to evaluate million to billion-member compound libraries within a single experiment at a fraction of the per compound cost, time, or material burdens of other hit-finding technologies.² Each small molecule within a DEL is tagged with an identifying DNA sequence and, through a multiplexed selection process, binder or active compounds within the DEL are identified for follow-up via high-throughput DNA sequencing. Many variations of this underlying concept have been developed, numerous examples of successful DEL syntheses and screens have been reported, and DEL screens have now become a regular fixture within many organizations’ target hit-finding plans.^3,4 Although the frameworks of DEL synthesis and screening have been reviewed, hit-to-lead optimization following DEL screens has received less attention.⁵ In this Innovations article, we review recently published hit-to-lead optimizations following DEL screening, observe trends, and examine key questions. What are the molecular properties of DEL hits which are selected for hit-to-lead optimization? What strategies are pursued during the optimization? How different or similar are they to practices employed optimizing hits from high throughput screens (HTS) and other approaches?

Characteristics of DEL

DEL exhibits many appealing features and, like any hit-finding method, has limitations and artifacts which affect compound library design, screening, and follow-up. Most DEL chemical matter was built through split-and-pool “DNA-recorded” synthesis using 2–4 cycles of combinatorial DNA-encoded chemistry, and each cycle may involve dozens to thousands of building blocks. The curation of large, diverse sets of chemical building blocks with tightly defined physicochemical properties can be difficult, leading to product structures that may span outside desirable property ranges. During synthesis, characterization is conducted mostly through mass spectrometry, and limited purifications are employed—thus, variable yields of expected structures, synthetic truncations,⁶ undesired side reactions, or unidentified byproducts may all be encoded by a single DNA sequence. Although DEL-compatible reactions are being continuously developed (e.g., photoredox-based couplings, reversible solid-phase DEL synthesis and C–H activation), historically DELs made heavy use of a small number of robust, aqueous reactions (e.g., amidation, sp² cross-couplings, nucleophilic substitution, and reductive amination).⁷ Due to the covalent connection to the associated DNA tag, individual DEL structures can exhibit linker-based effects during the screen and hits will often bear functional groups at the site of linker attachment containing hydrogen bond donors/acceptors that contribute to hit binding. In addition, due to the presence of an attached DNA tag, any very large or highly lipophilic structures formed during DEL synthesis can be expected to stay solubilized in water during screening. Within a DEL selection for binders, initial protein–DEL stoichiometry may be skewed to avoid binding site competition effects—however, binder identification is dependent on selection stringency and sequencing coverage, ultimately leading to preferential identification of higher affinity compounds.⁸ This is in contrast to other high-volume approaches such as HTS in which collections of compounds are screened as individual, nonlinkered structures prepared with no inherent limitations on synthetic chemistry or reaction sequence and with fewer limitations on assay formats or ranges of compound detection (Figure 1).^9,10 A detailed comparison of screening modalities is beyond the scope of this article, and readers are referred to ref (3) for an excellent overview.

Figure 1.

Comparison of DNA-encoded library and high-throughput screening technologies.

Although design methods to control DEL hit properties have been proposed,¹¹ complete descriptions of realized DEL productions and the properties of their enumerated structures and resulting hits remain scarce. However, several reviews have surveyed the physicochemical properties of hits arising from available DEL screen data. A comprehensive 2016 review from Franzini and Randolph catalogued 155 published DEL hits across 32 selections.¹² In their analysis, the number of diversification points used in the originating DEL tracked with increasing hit polar surface area, hydrogen bond donors, hydrogen bond acceptors, and rotatable bonds. While many hits contained cLogP values considered within druglike ranges, the majority of reported DEL hits had molecular weight above 500 Da, and nearly all were above 400 Da—properties considered to be outside of ideal leadlike space.^13,14 Similarly, a 2018 analysis on hits from DEL screens conducted at Roche revealed hits followed a normal distribution centered around 500 Da.¹⁵ Given these reported trends among DEL hits, several questions arise about the hit-to-lead optimization of these hits: (1) Do DEL screens, as they have been historically performed, explore a subset of traditional small molecule chemical space represented in HTS? (2) Do DEL campaigns deliver optimizable leads across target classes? (3) Are unique hit-to-lead tactics applied to DEL-derived hits versus HTS? (4) What molecular weight and lipophilicity ranges define the optimizable DEL hit space, and how does this differ from the total DEL hit space? (5) Is there a greater focus on defining the minimum pharmacophore, since a linker attachment point is known; and since truncated sublibraries are present in screening mixtures?

Molecular Properties of Optimizable DEL Hits and Changes during Post-DEL Hit-to-Lead

To address these questions, we identified peer-reviewed hit-to-lead optimization studies following DEL screens published within the past 5 years. Reports which presented compound data for DEL hits that were limited to on-DNA hit validation, only explored simple DNA-attachment point hit variants (e.g., acid, methyl amide), or only contained very narrow structure–activity follow-up were excluded from our analysis. After identifying 15 reports that passed these criteria, for each we determined structures of the hit and lead.¹⁶ Physicochemical parameters were calculated using a single consistent set of methods, and hit versus lead property changes were compared. The list of identified structures, their physicochemical properties, the target, and the target class are presented in SI Table 1 (see Supporting Information).

The hits contained in SI Table 1 represent the subset of total DEL hit space which has been shown to be optimizable. The hits arise from screens predominantly versus soluble intracellular targets, most commonly enzymes. The DEL library designs from which the hits originate vary, and just one example originates from a polypeptide library based on natural amino acids. The physicochemical properties of this hit and lead are outliers versus the other structures analyzed.

We compared the MW and cLogP of these optimizable DEL hits and their resulting leads using a chart based on that in two prior publications (Figure 2A).^6,12 Molecular weight is shown on the y-axis with 500 Da as a reference line, and clogP is plotted on the x-axis with reference lines at 1 and 5. DEL hits are shown as filled circles, with blue denoting hits which were optimized to leads and gray denoting hits whose optimization has not been published (gray filled circles are from ref (11)). DEL leads are shown as filled red circles. To add a reference data set, HTS hits are plotted in Figure 2A as purple open circles. Because public access to HTS hit structures is limited, we enabled this comparison using Janssen HTS data, for a collection of hits published previously.¹⁷

Figure 2.

(A) DEL hits and leads compared to HTS hits, MW plotted versus clogP. Optimizable DEL hits (blue filled circles), leads resulting from DEL hits (red filled circles), and entire DEL hit space (gray filled circles, ref (12)). Janssen historical HTS hit space (purple open circles). (B) Comparison of the MW between DEL hit and lead pairs. (C) Comparison of the clogP between DEL hit and lead pairs. (D) Comparison of lipophilic ligand efficiency between DEL hit and lead pairs. (E) Comparison of the ligand efficiency between DEL hit and lead pairs.

Figure 2A compares optimizable DEL hits (blue) and leads (red) with the total DEL hit space (gray) and Janssen HTS hit space (purple) across MW and cLogP. The total DEL hit space lies in a MW range which is higher than the HTS hit space (∼400–1100 vs ∼200–700 Da), but the cLogP range is similar. While the overall DEL hit MW range is higher (blue and gray in Figure 2A), the hits that were demonstrated to be optimizable (blue) generally fall in a lower MW range (350–700). This trend is unchanged when the HTS hit data set is filtered to only those hits which arise from screens versus enzymes (data not shown) On average, the optimizable DEL hits had MW 533 and cLogP 3.9, while their leads had MW 552 and cLogP 4.0 (see SI Figure 1).

Molecular weight and clogP changes between pairwise comparisons of hit and lead (Figures 2B and 2C) revealed changes in both increasing and decreasing directions and of variable magnitude. The property changes during DEL hit-to-lead are consistent with recent analyses of hit-to-candidate optimizations.^4,18,19 In contrast to MW and clogP, ligand efficiency and lipophilic ligand efficiency parameters showed a clear and consistent upward trend (Figures 2D and 2E). This suggests that a variety of tactics—truncation, MW growth, and increase or decrease in lipophilicity—consistently resulted in more potent leads with higher ligand efficiency.

An additional set of trends emerges from examination of molecular property distributions for HTS hits, all DEL hits, optimizable DEL hits, and DEL leads (SI Figure 1). Molecular weight and clogP distributions of all DEL hits span a significantly broader range than HTS hits, but the distribution of MW and clogP for optimizable DEL hits does not. The optimizable DEL hit distributions are narrower and similar to those for HTS hits. Interestingly, the MW distribution of optimizable DEL hits is centered at 533, intermediate from the midpoint of the HTS distribution (410) and total DEL hit space (631). Unlike MW, the midpoints of the clogP distributions are more similar (3.6–4.1).

Examples of Hit-to-Lead Optimizations

Changes in molecular properties during hit-to-lead campaigns based on DEL-derived hits are informative, and a more detailed examination of specific optimizations reveals what structural changes underlie the molecular property changes. In reviewing the published studies, a few trends emerge: (1) truncation of significant portions of the hit structure; (2) a focus on reducing lipophilicity; and (3) use of the DNA attachment vector to introduce polar functionality. These trends reflect commonly employed tactics in medicinal chemistry—minimum pharmacophore development and modulation of physicochemical properties—which are applied regardless of hit origin. However, aspects of the DEL hit optimizations diverge from those of HTS-derived hits, as the following examples illustrate.

ADAMTS-4

Iterative affinity screening of His-tagged ADAMTS-4 (aggrecanase-1) using a 4-cycle triazine DEL library provided a library subfraction which was amplified and sequenced.²⁰ The sequence enrichment data suggested clear directions for hit validation and SAR at each of the three triazine substituents as all chemotypes arising from 4-(aminomethyl)benzoic acid and a 2,6-disubstituted tetrahydroisoquinoline were highly populated and no preferred substituent was observed at the remaining triazine substituent which included the DNA attachment point (1a–d, Figure 3). This set of observations represents a rich SAR which emerged directly from the screen. While a similar data set can sometimes arise following an HTS for densely populated chemical clusters, regular observation of SAR trends and hit families has been noted as a common feature in DEL screens.²¹ In the case of compounds where binding affinity has been validated through off-DNA resynthesis and retesting, the SAR observed from DEL sequencing can qualitatively suggest immediate areas to focus hit optimization, a clear difference from hit-to-lead following HTS. In the case of ADAMTS-4, definition of the minimum pharmacophore was prioritized, and the cycle 1 substituent was truncated (1a [R = benzylamino] → 1b [R = H], Figure 3) based on the screening data. An increase in potency was observed, along with a large decrease in clogP (∼2.5 log units) and molecular weight (105 Da). Despite the improvement in molecular properties, polar groups were explored at this same position to improve solubility. Both anionic and cationic substituents retained high potency, consistent with the authors’ hypothesis that this vector, the former DNA attachment point, was exposed to solvent. The use of the DNA attachment vector to incorporate polar functionality and modulate molecular properties is a theme which is explored in multiple DEL hit-to-lead examples.

Figure 3.

Hit-to-lead optimization targeting ADAMTS-4.

The solubility-enhancing groups incorporated into ADAMTS-4 inhibitors significantly increased molecular weight, and further truncation was explored. In the fully truncated scaffold (2a, R = H) a ∼100-fold decrease in potency was observed, and potency could be regained through reintroduction of a polar group at the DNA attachment vector (2b, R = dimethylaminoethylamine). The resulting lead exhibits a balance among potency, lipophilicity, and molecular weight, and profiling of compounds in this series indicates selectivity versus other zinc metalloproteinases.

IDO1

Hit-to-lead optimization following an IDO1 (indoleamine 2, 3-dioxygenase 1) DEL screen followed some of the same trends.²² Selection data suggested no preference at the cycle 1 position within the hit series, and off-DNA resynthesis validated that truncation at this position and at the DNA attachment point was tolerated, providing compound 3 with moderate activity in a peripheral blood mononuclear cell (PBMC) IDO1 activity assay (Figure 4). Sub micromolar activity of the validated, truncated hit in a PBMC matrix is notable and supported further optimization efforts. Replacing the indole in the screening hit with indazole and separation of isomers improved potency by >10-fold (4), and the optimization effort shifted to improvement in pharmacokinetic properties. Despite the drop in molecular weight and clogP provided through truncation, 4 exhibited low oral bioavailability and dose escalation led to even lower observed bioavailability. Compound 4 exhibits low crystalline solubility in fasted state simulated intestinal fluid (FaSSIF, 18.4 ug/mL), and the authors hypothesized that improvement in FaSSIF solubility would boost bioavailability. Despite the knowledge that a linker for DNA attachment was tolerated at the indazole 3-position, the authors did not report placement of solubilizing groups at this vector; rather, a prodrug strategy was pursued and successfully improved oral bioavailability. Whether solubilizing groups at the DNA attachment vector would have led to retained activity or improved solubility is unknown, and this example illustrates that while a DEL hit always contains a vector where large polar groups are tolerated, it is just one of the options medicinal chemists have to optimize molecular properties.

Figure 4.

Hit-to-lead optimization for IDO1.

BCATm

DEL screening against mitochondrial branched chain aminotransferase (BCATm) differed from ADAMTS-4 and IDO1 in that selection data gave no clear direction for removal of an entire building block subunit of the structure.^23,24 Off-DNA resynthesis was performed with a methyl amide at the DNA attachment vector (5), which is commonly used to truncate the linker region. Enzyme inhibition activity was validated, and improvement in mouse pharmacokinetic properties was pursued to enable an in vivo proof-of-concept study. Replacement of the methylthiophenyl substituent with pyridine and other heterocycles enhanced potency while decreasing lipophilicity and molecular weight, and cellular activity was retained (6a, Figure 5). Truncation of the entire methyl amide substituent led to a loss of potency (6b), and small modifications including dimethyl amide (6c) were not tolerated, suggesting the DNA attachment vector was sensitive to structural modification. Insight into the potency SAR at this position was provided by an X-ray cocrystal structure of 6a in complex with BCATm. At the DNA attachment vector, the methyl amide makes separate H-bond donor and acceptor interactions, and the methyl group is oriented toward a small hydrophobic pocket, consistent with the SAR that larger substituents are not tolerated. The amide N–H bond is oriented toward solvent, suggesting that during the DEL screen the linker and DNA are accommodated with this trajectory. The X-ray structure corroborated the SAR that cis stereochemistry was preferred within the 1,3-diaminocyclohexyl unit. A takeaway from this study is that while the DNA attachment vector of a DEL-derived hit must by definition accommodate a large hydrophilic substituent, off-DNA hit optimization should also explore nonpolar substituents at this position. It is not assured that property modulation via polar substituents at this vector will be tolerated or beneficial.²⁵ In the end, property modulation via incorporation of pyridine at the benzimidazole 2-position provided improvements in mouse pharmacokinetic properties sufficient to enable successful measurement of BCATm inhibition in vivo following oral dosing.

Figure 5.

Top: hit-to-lead optimization for BCATm. Bottom: X-ray crystal structure of 6a bound to BCATm; the DNA attachment vector is indicated in the yellow box. (PDB code 5HNE).

BRD4

In a separate study, DEL screening identified a structurally novel second generation bromodomain and extraterminal (BET) inhibitor using His-tagged BRD4.²⁶ The hit-to-lead optimization strategy was derived from an analysis of attrition causes for the first generation BET inhibitors as well as the risk of submaximal target engagement for these compounds due to low systemic concentrations following oral dosing. The strategy therefore emphasized physicochemical properties, solubility, and predicted human daily dose. As with ADAMTS-4 and IDO1, sequencing data on the enriched library fraction following affinity screening immediately indicated that cycle 3 building blocks contributed little to binding. Once again, truncation was explored at the off-DNA resynthesis and hit validation stage, and the piperidine N-acetamide analog of 7 was selected as the starting point for optimization due to its favorable hit profile, in particular ChromLogD_7.4,²⁷ PFI (Property Forecast Index),^28,29 LE,³⁰ and LLE.³⁰ Notably, hit-to-lead progressed with deletion of the methyl amide at the benzimidazole 6-position, the DNA attachment vector, as this gave a further improvement in PFI. As a result of the extensive truncations, the starting point for off-DNA optimization had a molecular weight of 377, whereas the on-DNA structures detected in the DEL screen had molecular weights >500 Da. Replacement of the 2,6-dimethylphenol was prioritized to improve mouse pharmacokinetic properties, and a pyridinone isostere retained potency and improved PFI, LE, and LLE substantially. Incorporation of the pyridinone was suggested by the X-ray structure of a structurally related fragment bound to BRD4; the fragment originated from a screen unrelated to the DEL screen. The property-enhancing pyridinone was combined with further truncation of the piperidine to provide pyran 7a which, despite its low molecular weight, high solubility, and desirable physicochemical properties, had a predicted human dose outside the acceptable range. Potency enhancement, further improvements in pharmacokinetics, and maintenance of high solubility became the focus. While the dimethoxypropyl analog of pyran 7b was improved across these parameters, dose projections remained high. Therefore, the authors pursued substituents at the benzimidazole C5 and C6 positions. Incorporation of a polar substituent at C5 was suggested both because this was the DNA attachment vector and because, unsurprisingly, X-ray crystal structures indicated this position is oriented toward solvent. A number of N-linked amines were investigated and morpholine 8 provided sufficient improvement in both potency and metabolic stability to generate a >100× lower human dose prediction. This study provides additional evidence for the utility of exploring polar substituents at the DNA attachment vector; just as important, however, it shows that modulation of molecular properties was achieved through exploration of changes throughout the entire structure and not only at the DNA attachment point (Figure 6).

Figure 6.

Top: hit-to-lead optimization for BRD4. Bottom: X-ray crystal structure of 8 bound to the first bromodomain of BRD4; The DNA attachment vector is indicated in the yellow box. (PDB code 6TPZ.)

DDR1

Discoidin domain receptor 1 (DDR1) and its homologue DDR2 were subjected to DEL screening of pooled chemical libraries using a range of protein concentrations in an iterative protocol.³¹ A spirocyclic imidazolinone hit was prioritized based on selectivity for DDR1 over DDR2 and structural novelty among reported receptor tyrosine kinase inhibitors. Sequencing data from the screen did not support truncation of any building block units, and improvements in potency, solubility, and microsomal stability were pursued based on validated hit 9a, which contains a methyl amide at the DNA attachment vector (Figure 7). An X-ray cocrystal structure of DDR1-bound 9a indicated that this amide bound near an unoccupied hydrophobic pocket, and analogs which extend the methyl amide, such as the trifluoroethyl amide, improved potency ∼100×. As described above in the case of BCATm, the placement of a lipophilic substituent at the DNA attachment vector can be counterintuitive, since this vector by definition accommodates a large polar substituent. The DDR1 structure, however, demonstrates that an unoccupied lipophilic pocket can be found in proximity to solvent-exposed surfaces of the protein. Additional DDR1 SAR supports this observation, since morpholinoethyl amide 9c, which projects a polar substituent at the same position, boosts potency and decreases lipophilicity. A combination design exploring simultaneous occupancy of the lipophilic pocket as well as projection of a polar group toward solvent was successful, and N-hydroxyethyl, N-trifluoroethyl amide 9d maintained high potency and reasonable measured logD while increasing aqueous solubility >10-fold. This is a unique example of the DNA attachment vector being leveraged for both hydrophobic interactions and polar substitutions within the same compound. However, the property improvements came with a substantial increase in molecular weight, and human and mouse metabolic stability were poor versus the DEL hit (compare 9d and 9a). As in the BET hit-to-lead example, the authors explored reducing lipophilicity not only by leveraging the DNA attachment vector but also by incorporating polarity throughout the structure; in this case indazole → azaindazole substitution (10) provided the same potency, logD, and solubility improvements with a significantly lower molecular weight and improved metabolic stability. The resulting lead was further optimized to provide a tool suitable for in vivo proof-of-concept studies.

Figure 7.

Top: hit-to-lead optimization for DDR1. Bottom: X-ray crystal structure of 9a bound to DDR1; the DNA attachment vector is indicated in the yellow box. (PDB code 6FEW.)

Summary

We identified and analyzed recent published accounts of hit-to-lead optimization following DEL screens. A number of trends in molecular property changes and specific medicinal chemistry tactics were observed. Although in many respects the trends observed here are consistent with hit-to-lead approaches following HTS, there were also notable differences.

The nature of hit optimization is heavily influenced by the goals of the program and the properties of the hit—most studies we analyzed sought orally bioavailable leads; thus, the target property range was consistent with the well-described space of orally bioavailable small molecule drugs.¹⁸ The properties of DEL-derived hits, however, were significantly different from HTS-derived hits. A portion of DEL hits overlaps with HTS hit space (clogP vs molecular weight), but as a group DEL hits were higher molecular weight than HTS hits (see Figure 2A). This finding is consistent with the split-and-pool, molecular weight additive synthetic routes commonly employed in DEL synthesis. Notably, DEL hits and HTS hits span a similar distribution of clogP—a consequence of the wide array of polar and nonpolar building blocks that may be utilized in DEL builds largely unencumbered by difficult substrate-specific purifications.

Although DEL hit space partially overlapped with HTS hit space, the majority of DEL hits selected for optimization lay within the HTS hit space. This trend may result from selection bias, where lower molecular weight DEL hits are favored, or publication bias; in either case optimizable DEL hits arose disproportionately from this lower molecular weight range. In addition, the clogP of DEL hits selected for optimization lies within a narrower range than the overall DEL or HTS hit clogP range (∼1 to 6 versus −2 to 8, Figure 2A and SI Figure 1). Hit-to-lead optimizations of DEL hits emphasize truncations and identification of the minimum pharmacophore. This is a universal optimization strategy; however, it also aligns with the need for molecular weight reduction to achieve oral bioavailability from a high molecular weight DEL hit. Ligand efficiency emerged as the single parameter which universally increased during optimization, and lipophilic ligand efficiency had a strong increasing trend as well—despite variable changes in molecular weight and clogP across the studies. The LE and LLE changes appear to reflect the focus on truncation, potency enhancement, and an additional major trend, reducing lipophilicity. While optimizing the balance among potency, lipophilicity, and molecular weight is not unique to DEL derived chemical matter, aspects of how this was pursued in the DEL hit-to-lead studies reflected special considerations present in DEL. Lastly, the degree of structural modification from hit to lead varied and was, in some cases, considerable, as Tanimoto similarity between hits and leads ranged from 0.6 to 0.96. This is consistent with some optimizations introducing substantial core modification and not only varying substituents around a consistent scaffold.

Given the trends we observed, what should medicinal chemists consider when optimizing hits from DEL screens? (1) Leverage the SAR available from screening whenever possible and in particular to define the minimum pharmacophore. (2) Exploring the SAR of the DNA attachment point is productive and should include truncation, simple amides (primary, methyl, and dimethyl), and polar, nonpolar, and amphiphilic substituents. Consider acyclic tertiary amides simultaneously projecting a nonpolar and a polar group away from the scaffold. (3) The DNA vector may be oriented toward solvent and provide a handle to optimize physicochemical properties, but it is possible and worth considering that the DNA vector may point to a hydrophobic pocket which can be exploited for potency gains. (4) Leverage X-ray cocrystal and other structural information to inform designs at the DNA vector and throughout the structure. (5) If the goals of the optimization include oral dosing, pursue lower MW hits even if less potent.

A limitation of our analysis is that the protein targets in published DEL hit-to-lead studies represent only a narrow subset of target space being pursued in drug discovery.²¹ Enzymes are far overrepresented (13 out of 15 DEL hit-to-lead papers, versus 28% of drugs³³) and DEL hit-to-lead success rates and trends may differ for nonenzyme targets. Interestingly, DEL screens for other target classes have been reported,³⁴⁻³⁶ both for soluble (including nuclear receptors, structural proteins, and transcription factors) and membrane-bound proteins (GPCRs). But hit-to-lead optimizations based on these screens have not been published; it is unclear from the literature whether this reflects the optimizability of DEL hits for these targets. Targeted protein degradation is an emerging area with high potential for DEL impact, but studies describing evolution of DEL screening hits to degraders have not yet appeared. The outcome of hit-to-degrader optimizations, which lie in the beyond-rule-of-five property space, will likely differ from those in the traditional small molecule space.

As further refinements in DEL design strategy and newer DEL chemistries make their mark within DEL screening collections, the properties of DEL hits will evolve and so will hit-to-lead optimizations. Novel DEL-compatible reactions may shift DELs toward more targeted, smaller-sized 2- or 3-cycle libraries that are within a more traditional HTS chemical hit space. Furthermore, new advancements in the analysis of DEL selection data may allow better understanding of lower-enriched structure–activity relationships³⁷ or train predictive models³⁵ to enable prioritization of hits that have improved physicochemical properties. Continued innovation in DEL design and synthesis is likely, and the hits which emerge from novel libraries will continue to drive hit-to-lead and the invention of future clinical candidates and drugs.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.0c00615.

• Literature search strategy; method of analyzing papers; characteristics of hit-to-lead studies; molecular property distributions for DEL hits, DEL leads, and HTS hits; DEL hit and lead structures and physicochemical properties (PDF)

Author Contributions

The manuscript was written through contributions of all authors.

The authors declare no competing financial interest.